The global rise in temperatures due to climate change has brought attention to the need for advanced forecasting models to understand historical trends and predict future climate patterns. This study investigates global and regional temperature variations using historical climate data, focusing on the impact of greenhouse gas emissions across major sectors. AnLSTM auto encoder model was employed alongside traditional machine learning models such as Linear Regression, Gradient Boosting, and Random Forest to forecast temperature changes. The LSTM autoencoder demonstrated superior performance in global data analysis, achievingan accuracy of 98.48% and an F1 score of 92.31%. However, regional performance varied, with traditional models outperforming in some cases, particularly in Africa and the Americas. Sectoral analysis revealed agriculture and power industries as the largest contributors to emissions globally, with regional variations in sectoral impacts which lead temperature rise.The findings highlight the importance of incorporating tailored modelling approaches and integrating socio-economic variables for better climate forecasting.